Research on Hybrid Feature Selection Method in Intrusion Detection

In an open network, redundant or incompatible attributes of high-dimensional mixed features reduce the efficiency of network intrusion detection.In order to improve the response performance of intrusion detection system, this paper proposes a hybrid feature selection method.Rough set theory is used to description for intrusion detection feature selection.Information entropy and average weight are used to define importance of numeric features and character features.After generating a descending feature sequence, K-means clustering algorithm is used to evaluate the optimal feature subset.Simulation experiment is done in KDDCUP99.It shows that the method is effective to select feature subset and shorten the detection time.